Epilepsy is an often debilitating neurological condition affecting 3 million Americans and more than 50 million people across the globe. Despite several decades of excellent clinical, genetic and basic research and the existence of dozens of animal models and hypotheses, the mechanisms underlying human focal epilepsy are still not understood. To achieve the no seizures, no side effects goal of epilepsy research, we need to first answer a set of fundamental questions: how do focal seizures start, how do they spread, and how do they terminate? In particular, what roles do different subsets of neurons - inhibitory vs excitatory - play in the progression of human seizures? Intracranial electrocorticogram (ECoG) recordings in patients with intractable epilepsy are used to localize the brain region where seizures originate. ECoG signals represent the summed activity of thousands of neurons, and have revealed many important macroscopic features of seizures. However, many of the mechanistic predictions arising from animal models of epilepsy are at the level of individual neurons, and cannot be tested using ECoG alone. Here, specially designed recording techniques and devices are used to safely record the simultaneous activity of hundreds of individual neurons during seizures directly in patients with pharmacoresistant focal epilepsy. It i then possible to selectively identify human inhibitory neurons and ask how they control seizures. Many animal and slice studies state that decreased inhibition leads to seizures. However, many others state that increased inhibition is necessary to synchronize activity before a seizure can occur. Direct recordings of these inhibitory interneurons from humans present a unique opportunity to resolve this debate. By carefully identifying human inhibitory interneurons it is possible to characterize how they behave during all phases of human seizures. The activity of these human inhibitory interneurons can then be compared to that of different kinds of excitatory cells. The activity of inihibitory neurons can then also be manipulated optogenetically in mouse models of epilepsy to confirm that the human observations linking inhibitory neuron activity and seizure intensity are causal, and not just correlative. This can point the field towards novel pharmacological, surgical and predictive therapies for epilepsy that specifically target particular neuronal subtypes.